Zo Computer is Your Personal AI Cloud Computer 💻⚡

CV Deep Dive

Today, we’re talking with Ben Guo and Rob Cheung, Co-Founders of Zo Computer.

Zo is an intelligent personal cloud computer designed to act as a fully customizable digital workspace: part agent, part server that adapts to its user. Unlike traditional assistants or no-code tools, Zo gives every person their own AI-powered server that can host apps, run automations, integrate with personal data, and even spin up custom software on the fly. The goal is to give individuals professional-grade infrastructure for personal use, enabling them to build highly tailored systems on top of their own data rather than relying on generic, one-size-fits-all software.

In this conversation, Ben and Rob share how Zo was founded, the opportunity in giving individuals their own intelligent server, and their vision for a more user-owned future of the Internet.

Let’s dive in ⚡️

Read time: 8 mins

Our Chat with Ben and Rob 💬

Ben, welcome to Cerebral Valley! Given your background designing APIs at Stripe and Rob’s experience as a founding engineer at Substack, what led you both to team up and build Zo?

I’m Ben, founder at Zo. I met Rob over 10 years ago at Venmo – both of our first jobs out of college. We connected quickly as friends and stayed close along with much of the early Venmo team. We went our separate ways for a while. Rob was the founding engineer at Substack, and before that he worked on this AI-before-AI assistant called Fin, which used humans in the loop supported by AI systems. I worked at Stripe for a long time, we both were both in SF, then eventually we ended up back in NYC around the same time. We were at a party one night, and it struck me that it was finally the right time to start something with Rob. This was post a lot of exciting developments in AI in early 2023, when we first started talking seriously. I had just crossed eight years at Stripe, so it made sense to try something new, and it made even more sense to work with Rob. I’ve always wanted to, and we’d built side projects together before. In a lot of ways, it’s a dream job for me.

I’m Rob, and I’d echo a lot of what Ben said. We’re great friends, and if you’re going to pour a lot of time, effort, and mental energy into something, it really matters who you’re working with. That’s the biggest factor for me. I’ve always been into AI and have tracked it for years. We worked on Substrate before this, which was an infrastructure company. The impetus for that company came from some ML problems I was working on at Substack, where I realized how wildly inefficient model inference was at the time. The first version of the company grew out of that intuition.

How would you describe the concept of an "intelligent cloud computer" to the uninitiated developer, and why is now the right time for a personal server that hosts itself?

The question is why an intelligent cloud computer, what that is and why now. I’d say you have to back up a bit. Our basic thesis is that computers are the most flexible tools ever invented, yet the software we use is incredibly generic. It makes sense for big companies to build for millions of users, and they essentially have to split the difference and create these monolithic tools that nominally work for everyone but are ideal for no one. It ends up this way because building good software requires expert knowledge that takes years to develop, which makes truly custom software basically infeasible for most people. The “why now” is that AI changes this economic equation.

One way to put it is that most people’s experience with software is feature-driven. A feature is something a company builds for you, you use it, and it hides all the complexity. What’s happening now is that expert knowledge is available on demand – fast and cheap, and getting faster and cheaper. That shifts the primitive back to the computer itself, the genuinely flexible thing. 

That’s what Zo is: more or less an autonomous computer that you can talk to. The server part ties into this – we can get into it later – but the timing finally makes sense.

It’s always been clear that servers are useful. All companies run on servers. The hard part was that no normal person could operate one – no time, no expertise. That’s changing. Everyone should have a place that organizes the logic and data they own, powering downstream software that uses their data and their workflows.

Your target is builders who find Zapier too limiting but full AI IDEs like Cursor too intimidating. Who are the specific users finding the most value in Zo right now?

Zo is the best place to build personal software, which is a very specific kind of software. I predict most consumer-generated software in the future will take this shape – software built on top of your own data. There’s no good place for that right now. 

If you look at something like Cursor or Lovable, they don’t have any of your context. You’re working in more traditional software engineering projects, deploying them, and handling all this stuff you don’t really want to deal with. It’s a huge pain.

Tools like Zapier also only cover a small slice of your data, and connecting everything you care about into a setup like that isn’t the right fit. Through our own tinkering with AI over the last several years, we've realized there’s no good home for our personal software projects. Zo is unique because it’s a unified workspace – not just for building lightweight vibe-coded projects, but for working across your entire digital life.

We take the definition of “software” very broadly. It’s not just small websites or scripts. It’s automations, APIs you set up, or anything you want to run on your server. Zo can do it all, and operate it for you.

We’ve seen internal examples like Rob’s personal health site or users building full SaaS decks. What are some of the most creative or unexpected ways people are currently using their Zo computers?

We call ourselves a computer company, and the whole point is that it’s very horizontal. Your computer is shaped around who you are and becomes useful based on what you do. Our use cases are naturally wide. We have biology researchers doing pretty hardcore research, using it to organize their hypotheses and experiments, we have a yoga teacher who hosts her booking site on zo. You can just ask Zo to VPN into a university server, and you don’t have to figure out all the stuff that normally makes that process cumbersome. It’s just a computer, and a computer can do anything. Now we have a natural language interface that makes it way easier to leverage that flexibility of the computer – which is amazing for non-technical folks.

The health data example is emblematic of what we’re going for, where we should own our own data. A lot of our health information, along with things like Spotify history or Amazon purchase history, is siloed in services we use. I can’t really look at how I’m spending money on Amazon or understand what I’m doing on Spotify other than the UIs they give me. A related example – going deeper than my personal health dashboard – is that I started doing a genomics study with my own genome from 23andMe. It’s incredibly involved, and I couldn’t realistically do it myself without something like this.

Where you have to go out and grab all the research databases and look at the new work coming out about variants. With Zo, it organizes a clean database of my whole genome and can make ad hoc scripts using the computer to pull in all this instrumental data. I can see how my genetics might affect how I should think about getting better sleep, for example. The fact that we can answer those kinds of questions now is really cool, and there isn’t a great analogous place to do something like that today.

Maybe the closest is something like Manus, but it’s not really designed for anything specifically about you. With my genomics question for example, I have a question that is very ad hoc, but it also has an async component. I also have a website that displays my genetics goals and related information. So I can ask questions actively, and I can also have this background passive layer running on top.

Walk us through the "Connect, Create, Deploy" workflow in Zo's unified workspace. If a new user wants to connect their Gmail or Linear to start building, how streamlined is that process?

It’s super streamlined for getting started with Zo, and you can connect as much or as little as you want. We really recommend connecting text so you can text your computer, which is surprisingly fun. You can also email your Zo's email address, which is another unique way to interact with your AI computer. Beyond that, we have dozens of built-in tools including Gmail, Linear, Notion, Google Calendar, Dropbox, Airtable, and Spotify.

The cool thing is that because Zo is an AI running on your server, if there’s anything that isn’t supported, you can just ask Zo to build the integration for you. Zo will walk you through getting the API key and setting it up, even if you’re not technical. This resembles some of the recent AI systems like Poke, where you can connect different things and have it do tasks for you. But the difference with Zo is that it’s far more customizable. The ceiling is way higher in terms of how deeply you can change everything about the system, because it's your own server. You can set up a complex health-tracking system or a CRM on top of Zo, and it will work exactly the way you want with the trigger words you prefer. Even the Zo application itself runs on your server, and we plan to allow full customization – or even replacement – of your Zo's interface in the future.

The sky’s the limit in terms of what’s possible when you’re customizing your own server that you fully control.

https://x.com/zocomputer/status/1991168904551436485?s=20

How are you measuring the success of a platform designed for personal software ownership? Are there specific engagement metrics or deployment milestones you are most focused on?

For us it’s daily active users. I’d say the whole premise is that this thing should shape itself to how you live, and the measure of success is whether people adapt it to themselves in a way they find consistently useful. A system like this does take work. It’s like buying a house – you have to invest in it. You have to furnish it and make it the way you want. But once you have it set up, it becomes this invaluable thing you can’t live without. So daily active users is how we measure ourselves.

There are many AI workspaces like NotebookLM and app builders like Replit, but you explicitly position Zo as a "personal server." What sets Zo apart from those tools from a product perspective?

So the way I see this space is that there are agents that act more like personal assistants – things like Poke – and there are AI tools that let you build – things like Replit. Zo is all of them in one, and it has all your context. It also has built-in hosting because it’s your server.

Building software in Zo is more ergonomic simply because everything you care about lives in one place. And the fact that it’s a server – just to clarify the Reddit differentiation – makes working with applications you build far more intuitive. It’s much simpler for a non-technical audience to understand where their data is and what’s happening to it. You don’t have to think about deployment at all, because your development environment is your runtime environment, and you're just working with files on your server. I think it's a much more intuitive model for software development, to simply build in the same place it runs.

It’s much easier than dealing with all the traditional software engineering baggage that present-day vibe coding tools are carrying from the old world. When we imagine how this new wave of builders will build software, we should ditch the vestigial organs.

You’ve built on the shoulders of giants like Modal, Steel, and PydanticAI. Could you share how Zo works under the hood and the reasoning behind your architectural choices?

I’d say there are a number of interesting architectural and software problems involved in giving people a system like this. I’d probably start with the fact that we’re in a very fortunate time in history to be building something like it. We're defining a new computer category for consumers – it’s a personal server – but what we’re actually riding on is the last 15 to 20 years of incredible enterprise compute. The whole industrial compute stack has been made efficient and reliable because of all the SaaS and large-scale software that’s been built. Our goal is to create the same level of professional-grade infrastructure for very personal use.

We can use data centers and undersea cables, but we’re also building on top of the container tech developed over the last 15 years, the way modern systems deploy software, and how hardware and software are separated. When a normal consumer thinks about having a cloud computer, it’s very different from having a laptop. In these environments the hardware is completely separate from the software, which lets you do things your laptop could never do. You can instantly switch from having 7 CPUs to 64 CPUs, or have a supercomputer-level setup within seconds.

And I can use that for 10 minutes, get charged about 7 cents for it, and then scale back down. That’s something a normal computer user wouldn’t expect. My MacBook Pro is a $5,000 machine, and most of the time it’s just doing email or sitting idle. So we’ve thought a lot about how to leverage the industrial software side of things to give people something much more dynamic.

There are other good reasons to do it this way. Another thing consumers aren’t used to with their laptops is being able to roll back their entire computer to a previous point in time. This is very important in the AI world. Because of container tech – the layering, the efficiency of storage – we can snapshot constantly. If your AI messes something up, deletes a file, or does something destructive, you always have the option of system-level time travel. You can’t do that on a laptop, because the architecture just isn’t built for it.

A lot of the feeling of having this “magical server” comes from fairly sophisticated systems work – the kind you’d normally do at a massive-scale product like Substack – but adapted for individual consumer use.

Since Zo handles everything from file storage and context management to actual hosting, what has been the hardest technical challenge in building such a cohesive platform?

A lot of it has to do with actually getting the complex systems work to function reliably. It’s a different kind of challenge because we give people total root access to their machines. That’s not a pattern you see in normal software startups. One of the big problems is that as a platform computer company, we have to manage this very "live" system with what the Linux kernel calls user space. There’s user space and platform space, and you don’t actually know what someone is going to do in user space.

A big part of what we have to balance is updating the system underneath all these live instances while improving the platform, all while users can do arbitrary things. With total root access, they can delete everything or change anything they want. Balancing that interaction is a pretty new software and technical problem.

Before we move back to the vision, are there any other technical implementation details that you'd like to share?

I’d say another big piece is networking. Once you have a server on the internet, you suddenly have a permanent home out there for the first time. We all have smartphones and laptops, but they’re on and off, and they mostly just read from the internet. Now you actually have something that exists continuously – just like Google has servers, of course on a much smaller scale – and being able to communicate with it anytime, from anywhere, is a major part of the value.

We take on the networking across all the protocols and ways you might want to interact with your server. You can host a database, or a Minecraft server, or something running over raw TCP. You might even set up a P2P protocol like BitTorrent between two Zo hosts. Making the networking transparent and simple for non-technical people is a big and interesting problem.

You’re on a mission to make the internet "wild and free again" – how do you foresee Zo evolving over the next 6-12 months to get closer to that vision?

The starting point is making it really easy to get your personal data from various places into Zo. Your personal data is already widely available because of GDPR. You can do a Google Takeout, you can download it from Amazon, and we want to make it really easy to use prebuilt templates to build automations, or dashboards, or anything you want on top of that data. The big goal is to enable this "true citizenship on the Internet" vision that we talk about. What we’re envisioning is a more distributed, decentralized future for the Internet, where every individual participant owns their own server.

What we want to enable over time is the ability for people to interact, create communities, and form networks that are more peer-to-peer than the centralized world we live in today. When you own a server, it becomes really easy to go server-to-server. And because your server is intelligent, that means it can autonomously interact with another Zo server. And then you have autonomous commerce, and everything that will bloom on top once you have that ecosystem.

Tell us a bit about the team culture you’re building at Zo are you currently hiring, and what do you look for in prospective engineers?

That’s an interesting question because engineering is so different now. I worked eight years managing teams, and the profile you look for has changed. We’re looking for people who are very AI-pilled in their programming – people who have fully adopted the tools, are curious about how to use them effectively, and are constantly optimizing their workflows. Because of what Zo is, those people naturally have good instincts about what to build and what makes sense ergonomically. We’re looking for engineers who are very pro-AI in their coding workflows and have big visions for what the future of engineering looks like, which will look very different – working at a much higher level of abstraction.

Rob can also talk about the exact profile we’re looking to hire.

Yeah. The other half of it is deep competence. We’re a systems company. One of the main roles we’re hiring for now is a founding infrastructure engineer. I’ve been doing most of it myself with a small, lean team, but AI or not, deep competence in systems takes a lot of context. It takes a lot of war stories you just have to have lived through. The other side of how we think about hiring is that companies are built on great teams, and having a lot of effective, competent people is what makes some products truly exceptional.

We’ve got investors including Lightspeed, South Park Commons, Craft Ventures, Guillermo Rauch from Vercel, Immad Akhund from Mercury, and a host of smaller angels. We raised for Substrate, the original inference platform we mentioned, back in 2023.

Anything else you'd like our readers to know about your hard launch or how to get involved with the Zo community?

I’d say we’re still early but excited about this launch. We’ve been in a small public beta and have a pretty active community. We have people who are genuinely obsessed with Zo. They've canceled their subscriptions to other products like ChatGPT, or Squarespace, or Zapier for Zo, because the ceiling is just so much higher, and it has so much range in terms of all the things it can do. We have lots of people using Zo for many hours a day for their core workflows, which is really exciting. One non-technical founder has adopted Zo as his personal CRM and set up this wild system with transcription and personalized follow-up emails that blew me away the first time he showed it to me.

The best place to get involved in the community is Discord, and we’re really excited about building this into a broader community. The vibe we’re trying to evoke is the Homebrew Computer Club that birthed the first Apple computer, but for the modern age. We’re in this really unique moment where everyone is exploring new tools, and we have an opportunity to rethink how we interact with computers and with the Internet.

Our vision for the community is really much bigger than the Zo product. We just want to gather like-minded folks who are excited about the moment we're in – people who just want to share the latest cool thing they built with AI for themselves, or compare notes on new AI models and tools. We talk a lot about how our mission for the company is to "teach people how to fish" with computers. AI is leveling the playing field, it's not about learning to code anymore, but you still need to understand some concepts – it's like a much more accelerated curriculum is possible, and we believe an intelligent server is probably the best place to learn that curriculum.

Conclusion

Stay up to date on the latest with Zo Computer, follow them here.

Read our past few Deep Dives below:

• Arcade - A New Kind of Marketplace 🔍

• Strong Compute - Your High-Performance Training Infra Platform 🔋

• V7 Go - AI Agents for Complex Document Workflows 🎛

• Zencoder’s Repo Grokking - Your AI coding unlock 🔑

• Rime - Authentic AI Voices for Real-time Conversations at Scale 🗣️

If you would like us to ‘Deep Dive’ a founder, team or product launch, DM our chatbot here.

CV • Instagram • X • All Events